Accurate and reliable location information is the basis for the autonomous navigation of off-road vehicle. In this paper, a multi-sensor navigation positioning system, integrated with a RTK-GPS (Trimble AgGPS332/MS750) and a CCD camera (OK AC1310) was constructed. The RTK-GPS was used for obtaining the absolute position data, heading angle and speed of the vehicle; and the calibrated CCD camera was used for obtaining the baseline for navigation through the image processing and the feature points of crop rows. Based on the kinematics model of navigation vehicle, UKF filter was established to filter the information of the two sensors. An electric power vehicle was applied to the platform, and the position effects before and after filtering were compared. Experiments results showed that the standard deviations of X and Y were 2.44 m and 0.07 m separately with smoother location curves and the impact of jump points reduced after filtering by UKF. This multi-sensor navigation positioning system can overcome the disadvantages for using a single sensor and meet the requirements of the autonomous navigation system of off-road vehicles.